Dietary patterns and longevity

Nutrition research spent decades chasing individual nutrients, vitamin E, beta-carotene, dietary fat, and repeatedly came up short. The field has since shifted its unit of analysis to whole dietary patterns, and the evidence is considerably more coherent there. Several signals now meet a high evidentiary bar.

Mediterranean Diet: The RCT Evidence

The PREDIMED trial assigned roughly 7,400 high-cardiovascular-risk adults to one of three arms: a Mediterranean diet supplemented with extra-virgin olive oil, a Mediterranean diet supplemented with nuts, or a low-fat control diet, followed for a median of five years. Both Mediterranean arms showed approximately a 30% reduction in major cardiovascular events (heart attack, stroke, or cardiovascular death) compared with the control [1]. One disclosure is worth making: the original 2013 publication was retracted and republished in 2018 after investigators identified a cluster-randomization error. The corrected analysis held the conclusion intact, but readers deserve to know the history.

PREDIMED established CVD event reduction. The CORDIOPREV trial extended this to secondary prevention, following 1,002 patients who already had coronary heart disease for seven years. The Mediterranean diet arm showed a statistically significant reduction in carotid artery wall thickness and plaque regression; the low-fat arm did not [2].

On all-cause mortality, the best available estimate comes from observational data. A meta-analysis by Sofi et al. found that each 2-point improvement in Mediterranean diet adherence score corresponded to an 8% lower risk of dying from any cause (RR 0.92), alongside reductions in cardiovascular disease, cancer, and neurodegenerative disease [3]. This is prospective cohort data, not an RCT, so it complements PREDIMED rather than confirming mortality benefit on its own.

MIND Diet and Cognitive Aging

The MIND diet combines Mediterranean and DASH eating patterns, with components selected specifically for brain health. Researchers evaluated it in the Rush Memory and Aging Project: 960 older adults followed for a mean of 4.7 years. Adults in the highest tertile of MIND adherence showed a rate of cognitive decline roughly equivalent to being 7.5 years younger than those in the lowest tertile, across five cognitive domains. The association held after adjusting for APOE ε4 status, cardiovascular risk factors, and depression, and a dose-response gradient appeared even at moderate adherence [4]. A companion analysis found approximately 53% lower Alzheimer's disease incidence in the highest-adherence group. That finding rests on a smaller number of incident cases and deserves more caution until it has been replicated [4].

Both findings are prospective observational. The MIND-ADMPC RCT will eventually test whether the cognitive protection survives randomization. Until those results arrive, the observational consistency is suggestive but not confirmatory.

Ultra-Processed Foods and Mortality

Two large prospective studies from the NutriNet-Santé cohort quantify the association between ultra-processed food (UPF) consumption and health outcomes. In 105,159 adults followed for a median of 5.2 years, each 10-percentage-point increase in the proportion of ultra-processed foods in the diet corresponded to a 12% higher risk of overall cardiovascular disease (HR 1.12, 95% CI 1.05-1.20), with similar hazard ratios for coronary heart disease and cerebrovascular disease [5]. A separate analysis of 44,551 adults followed for 7.1 years found that the same 10-point UPF increment was linked to approximately 14% higher all-cause mortality [6]. Critically, both associations persisted after adjusting for overall dietary quality scores and nutrient profiles, suggesting that the NOVA ultra-processed classification captures something beyond simple nutrient composition, possibly food additives, matrix disruption, or hyperpalatability.

One limitation applies to both: these studies draw from the same volunteer cohort, which skews health-conscious. Effect sizes in less-selected populations may differ.

Dietary Fiber and Whole Grains

Few dietary associations are as well-replicated as fiber and whole grain intake versus mortality. A Reynolds et al. systematic review and meta-analysis of prospective studies and RCTs found that higher fiber and whole grain intake produced dose-dependent reductions of roughly 15-30% in all-cause mortality, cardiovascular mortality, coronary disease, type 2 diabetes, and colorectal cancer. The authors judged the relationship possibly causal [7]. A separate dose-response meta-analysis of 45 prospective studies found that 90 grams per day of whole grains corresponded to approximately 17% lower all-cause mortality, 22% lower cardiovascular mortality, and 15% lower cancer mortality [8]. These two analyses converge through overlapping but distinct mechanisms, and the whole-grain finding captures a food-matrix effect that goes beyond isolated fiber alone.

Protein: Source Matters More Than Quantity Alone

A NHANES-based analysis by Levine et al. found that adults aged 50-65 consuming high protein (20% or more of calories) had approximately 74% higher all-cause mortality and roughly four times higher cancer mortality compared with those consuming low protein (under 10% of calories) [9]. The cancer mortality excess disappeared when protein came predominantly from plants, implicating animal protein and IGF-1 signaling rather than protein per se. A counterpoint from the same analysis: after age 65, low protein intake was associated with increased mortality, reversing the mid-life pattern.

Several caveats apply to [9]: the data are observational, dietary assessment relied on a single 24-hour recall, the cancer mortality hazard ratio of approximately 4× has not been replicated at that magnitude elsewhere, and lead author Valter Longo has disclosed relevant conflicts of interest.

The animal-versus-plant protein distinction is further clarified by Song et al., who used the Nurses' Health Study and Health Professionals Follow-up Study (n=131,342) to model what happens when 3% of dietary energy shifts from animal to plant protein [10]. Replacing processed red meat showed the largest benefit (HR 0.66, or 34% lower all-cause mortality); replacing unprocessed red meat and eggs also showed benefit (HR 0.88 and 0.81, respectively). Replacing fish or poultry showed no significant mortality change, a nuance that undermines blanket claims that all animal protein is harmful.

Saturated Fat: The Replacement Is the Point

A Cochrane review of 15 RCTs covering roughly 59,000 participants found that reducing saturated fat reduced composite cardiovascular events by approximately 17% (RR 0.83, 95% CI 0.70-0.98), but only in trials where the replacement fat was polyunsaturated [11]. All-cause mortality showed no significant reduction in the RCT meta-analysis. Replacing saturated fat with refined carbohydrates attenuated or eliminated the cardiovascular benefit. The public health message about saturated fat has often omitted the replacement macronutrient, which is precisely where the evidence sits.

Where the evidence is contested

Protein quantity in mid-life: The Levine 2014 cancer mortality finding (HR approximately 4×) is the most contested figure on this page. It derives from a single observational dataset with a single dietary recall and has not been replicated at comparable magnitude. Longo's disclosed conflicts of interest add reason for caution. The directional signal, that high animal protein in mid-life may carry mortality risk, is consistent with the Song 2016 data, but the effect size in [9] should not be taken at face value until Mendelian randomization or RCT evidence either confirms or revises it.

Saturated fat and all-cause mortality: The Cochrane review ([11]) shows cardiovascular event reduction when saturated fat is replaced with PUFA, but no significant all-cause mortality reduction in RCTs. Some researchers argue this reflects insufficient trial duration or statistical power; others argue the mortality null finding is the correct signal. The evidence does not currently support the claim that cutting saturated fat extends life, only that replacing it with PUFA reduces cardiovascular events.

MIND diet, observational versus RCT: The 7.5-year cognitive-aging estimate from [4] is striking, but cohort studies of dietary patterns are vulnerable to healthy-user confounding that adjustment for known covariates only partially addresses. The pending MIND-ADMPC trial results will be the most informative nutrition-brain finding in years either way.

References

1. Estruch et al. 2018 (PREDIMED), NEJM — Mediterranean diet & CVD · PubMed 29897866
2. CORDIOPREV — Mediterranean diet 7-yr RCT · PubMed 34372670
3. Sofi F, Abbate R, Gensini GF, Casini A. 2014 (Am J Clin Nutr) — Updated meta-analysis Mediterranean diet adherence; 16 prospective studies;… · PubMed 24476641
4. Morris MC, Tangney CC, Wang Y et al. 2015 (Alzheimer's & Dementia) — MIND diet score (0–15); n=923, age 58–98y, mean 4.7yr follow-up; top tertile… · PubMed 26086182
5. Srour B, Fezeu LK, Kesse-Guyot E et al. 2019 (BMJ) — NutriNet-Santé; n=105,159; 5yr follow-up; UPF as % of dietary weight; HR for CVD events 1.12… · PubMed 31142457
6. Schnabel L, Kesse-Guyot E, Allès B et al. 2019 (JAMA Intern Med) — NutriNet-Santé; n=44,551; UPF % → ACM ~14% higher per 10% increment (HR ~1.14);… · PubMed 30742202
7. Reynolds et al. 2019, Lancet — dietary fibre & health outcomes · PubMed 30638909
8. Aune D, Keum N, Giovannucci E et al. 2016 (BMJ) — Dose-response meta-analysis of whole grain intake; 45 prospective cohorts; n=778,076; 90g/day… · PubMed 27301975
9. Levine ME, Suarez JA, Brandhorst S et al. 2014 (Cell Metab) — NHANES I; n=6,381, 18-yr follow-up; moderate/high protein (≥20% kcal) in ages 50–65:… · PubMed 24606898
10. Song M, Fung TT, Hu FB et al. 2016 (JAMA Intern Med) — NHS + HPFS; n=131,342; animal protein 10% kcal → plant protein: HR 0.66 ACM (0.59–0.75) in… · PubMed 27479196
11. Hooper L, Martin N, Abdelhamid A, Davey Smith G. 2020 (Cochrane Database Syst Rev) — 15 RCTs, n=56,675; reduced/modified dietary fat; saturated fat… · PubMed 32827219